An Arabic-first, AI-assisted farm-to-market coordination platform built for the Ministry of Climate Change and Environment. Buyers place advance produce demand, an AI translator turns it into farmer-ready instructions, and a rule-based engine pools nearby farms and allocates orders across them - all confirmed before harvest. One order. Many farms. Confirmed before harvest.
A buyer pastes a messy request. The AI translator reads it, returns a structured demand table with interpretation and confidence, and tracks every order from request to delivered - alongside live local-sourcing, shortfall, and waste-risk metrics.
Farmers declare what they can supply, see their farm profile and open harvest windows, and ask the Arabic copilot for advice - pooling, grade, packaging, and timing - the same screen, captured here in both light and dark.
The operator console runs the allocation engine on the demand pool - splitting each order across farms, flagging shortfalls, proposing backups and substitutes, and scoring fulfillment risk. Shown here fully in Arabic, right-to-left.
From a buyer's plain-language request to a confirmed, multi-farm harvest plan - an AI translator, a deterministic allocation engine, a risk model, and an Arabic-first copilot, working end to end.
Buyers type what they need in plain language. A rule-based natural-language parser turns it into a structured demand table - crop, grade, quantity, packaging, delivery window - and returns an interpretation, a confidence score, and the reasoning behind every line.
A deterministic solver matches each order to farms by grade fit, supply confidence, location, and distance - then splits a single order across Al Akhdar, Desert Leaf, and Oasis Fresh when no farm can cover the volume alone, surfacing exact shortfalls as it goes.
Every allocation is scored for fulfillment risk with written reasons and concrete mitigations, backed by automatic backup-farm fallbacks and a substitute advisor that proposes alternative produce when supply runs thin.
An Arabic-native copilot advises farmers on whether to join a pool, which grade and packaging to target, and when to harvest - referencing live open demand so the guidance maps to what the market is actually asking for.
Once an order is allocated, each farm receives harvest-instruction cards plus a quality-and-packaging guide tailored to vegetables, fish, or honey - turning an abstract market signal into farmer-ready actions.
Each batch carries a QR-style passport with end-to-end traceability and a local-sourcing tracker, so every delivery can be traced from a specific farm back to the original buyer demand.
An Arabic-first farm-to-market coordination platform: an AI translator turns free-text demand into farmer-ready specs, and a rule-based engine pools nearby farms and allocates orders across them.
Built by Team Kanban, the studio's competition and experimental build team, at University Hackathon: Farm to Market Innovation Sprint 2026 - 2nd place, with AED 8,000. See the full competition record.
